From 934ea0ef4338f00c797500b10c39f03b3cfc1692 Mon Sep 17 00:00:00 2001
From: Andre Bogus <bogusandre@gmail.com>
Date: Tue, 24 Oct 2023 21:53:54 +0200
Subject: [PATCH] fix another wasm32 problem

---
 Cargo.toml       | 2 +-
 README.md        | 2 +-
 src/simd/wasm.rs | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/Cargo.toml b/Cargo.toml
index 0567bb1..1b24de7 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -3,7 +3,7 @@ authors = ["Andre Bogus <bogusandre@gmail.de>", "Joshua Landau <joshua@landau.ws
 description = "count occurrences of a given byte, or the number of UTF-8 code points, in a byte slice, fast"
 edition = "2018"
 name = "bytecount"
-version = "0.6.6"
+version = "0.6.7"
 license = "Apache-2.0/MIT"
 repository = "https://github.com/llogiq/bytecount"
 categories = ["algorithms", "no-std"]
diff --git a/README.md b/README.md
index eca7dcf..ee028de 100644
--- a/README.md
+++ b/README.md
@@ -12,7 +12,7 @@ The [newlinebench](https://github.com/llogiq/newlinebench) repository has furthe
 
 To use bytecount in your crate, if you have [cargo-edit](https://github.com/killercup/cargo-edit), just type
 `cargo add bytecount` in a terminal with the crate root as the current path. Otherwise you can manually edit your
-`Cargo.toml` to add `bytecount = 0.6.6` to your `[dependencies]` section.
+`Cargo.toml` to add `bytecount = 0.6.7` to your `[dependencies]` section.
 
 In your crate root (`lib.rs` or `main.rs`, depending on if you are writing a
 library or application), add `extern crate bytecount;`. Now you can simply use
diff --git a/src/simd/wasm.rs b/src/simd/wasm.rs
index 712f4da..3351a2c 100644
--- a/src/simd/wasm.rs
+++ b/src/simd/wasm.rs
@@ -39,10 +39,10 @@ unsafe fn sum(u8s: v128) -> usize {
     let u16s = u16x8_extadd_pairwise_u8x16(u8s);
     let u32s = u32x4_extadd_pairwise_u16x8(u16s);
     let (u1, u2, u3, u4) = (
+        u32x4_extract_lane::<0>(u32s),
         u32x4_extract_lane::<1>(u32s),
         u32x4_extract_lane::<2>(u32s),
         u32x4_extract_lane::<3>(u32s),
-        u32x4_extract_lane::<4>(u32s),
     );
     ((u1 + u2) + (u3 + u4)) as usize
 }